In addition to the basic algorithms like Q-Learning, I also included slightly more advanced ones like the Successor Representation. The tool lets you see the successor maps for each state as the agent moves through the environment, which I think makes them much more intuitive!

I spent some time last week vibe-coding "The Beginner's RL Playground" a simple web tool for exploring many basic concepts in reinforcement learning. Users can adjust all sorts of parameters of the algorithm and environment and see the effects of those changes in realtime.

my recent headspace

A little late, but here is my list of books read in 2024. I am going to *try* to get back into more fiction this year, but it could just as easily tip the other direction and end up being all obscure psychoanalytic texts instead...

Wow, I was not expecting this rate of progress for LLMs on ARC-AGI. From effectively 0% to greater than human level in less than a year...